Nanoscale Engineering for Novel Computation Using Evolution
Hajo Broersma, Faustino Gomez, Julian Miller, Mike Petty and Gunnar Tufte
Living systems are able to achieve prodigious feats of computation with remarkable speed and efficiency (e.g. navigation in a complex environment, object recognition, decision making, and reasoning). Many of these tasks have not been adequately solved using algorithms running on our most powerful computers. When one examines the physical structure of living systems it is very difficult to understand how indeed such systems are able to perform computation. Two things stand out. First, the simplest living systems have a complexity and sophistication that dwarfs man-made technology. Secondly, these living systems were not designed specifically to do computation. In fact, as Richard Dawkins made clear in his book “The Blind Watchmaker” they were not designed at all. Instead they have been subjected to a long process of Darwinian evolution. Why is it that such a blind process can find solutions to these difficult computational tasks, when our best attempts at the principled design of solutions to computational problems are poor by comparison? It is because natural evolution is a bottom-up process that is able to exploit the emergent physical and chemical properties of molecules. It is par excellence a physical exploitation process. Natural evolution uses the enormous complexity of molecular interactions and also the huge parallelism of physical systems. This is the inspiration for the NASCENCE project which aims to emulate Nature and use computer-controlled Darwinian evolution to create sophisticated information processing systems in materials. The benefit of this approach to computation is that we can use the high parallelism and computational complexity of physical components to solve problems that are either computationally intractable or very difficult to model.
The technological drive to produce ever-smaller devices (Moore’s Law) is leading to the construction of machines at the molecular level. However, the basic computational paradigm is still Von Neumann. Molecular and nanoscale electronics concerns building molecule-sized analogues of transistors and assembling them into designs that are facsimiles of today’s solid state circuits. There are fundamental issues with this approach: molecules cannot be controlled and manipulated in the same way as silicon wafers, and it is unclear that macroscopic top-down design methods are appropriate. The NASCENCE project will use computer-controlled manipulation of physical systems, such as networks of nanoparticles, arrays of carbon nanotubes and films of graphene, to evolve them towards doing useful computation and information processing. This is physical computation in a dish.
Keywords: Unconventional computation, evolving nanosystems, evolutionary computation, genetic algorithms, information processing, hybrid systems, simulation, mathematical modeling.